Descriptions:
This tutorial from Michael Automates demonstrates a complete end-to-end workflow for using Claude to generate, combine, and backtest trading strategies derived from TradingView’s open-source indicator library. The creator downloads Pine Script source code from popular TradingView indicators, organizes them in a custom backtesting engine with a specific file-naming convention, then prompts Claude to systematically test indicator combinations across historical price data for the top ten crypto assets across 1-day, 4-hour, and 1-week timeframes.
The featured result — a strategy the AI labeled A2D, centered on kernel regression breakout logic — was selected by Claude as the optimal balance of return and drawdown, outperforming the baseline portfolio by 3.6x in median-adjusted terms across all assets. The video explains why Claude’s selection criteria prioritize drawdown control over raw profit maximization, and why median performance is a more reliable benchmark than average when outlier assets like Solana (which showed a 90,000% gain in backtesting) can distort results.
Practical setup details are covered throughout: how to source open-source indicators from TradingView’s public library, the folder structure and file-naming conventions the backtesting engine requires, how to configure Claude notifications for approval prompts during long-running sessions, and how to request additional price data from the internet when the bundled dataset is insufficient. A downloadable prompt document and the backtesting engine itself are offered via the video description. No coding skills are required to run the workflow.
📺 Source: Michael Automates · Published April 17, 2026
🏷️ Format: Workflow Case Study
